Vals AI MedCode: leaderboard
Models assign ICD-10-CM primary and secondary diagnosis codes to de-identified hospital records, each a discharge summary plus the progress and consult notes from the stay. Two certified coders annotated the 2,755 reference codes, and model output is scored against them, with breakdowns by primary versus secondary code and by ICD-10 chapter. Percentage correct, higher is better.
Metric: Accuracy (%). Source: www.vals.ai. Source: benchmarklist.com. Saturation forecast: Around 2030. 107 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 (Max) | 63.57 | #12 (Claude Opus 5) |
| 2 | Gemini 3.1 Pro (Preview) (High) | 59.06 | #58 (Gemini 3.1 Pro (Preview)) |
| 3 | Gemini 4 Argon (High) | 58.8 | #8 (Gemini 4 Argon) |
| 4 | Claude Fable 5 (Max) | 56.07 | #16 (Claude Fable 5) |
| 5 | Gemini 3 Flash (Preview) (High) | 55.92 | #82 (Gemini 3 Flash (Preview)) |
| 6 | Gemini 3.5 Flash (High) | 55.83 | #60 (Gemini 3.5 Flash) |
| 7 | Claude Opus 4.7 (Max) | 54.86 | #49 (Claude Opus 4.7) |
| 8 | Claude Fable 5.1 (Max) | 53.51 | #9 (Claude Fable 5.1) |
| 9 | Gemini 3.7 Flash (High) | 53.39 | #23 (Gemini 3.7 Flash) |
| 10 | Claude Opus 4.8 (Max) | 53.22 | #37 (Claude Opus 4.8) |
| 11 | Gemini 3.6 Flash (High) | 53.15 | #51 (Gemini 3.6 Flash) |
| 12 | Claude Sonnet 5.5 (Max) | 52.92 | #11 (Claude Sonnet 5.5) |
| 13 | GPT-5.1 (2025-11-13) (High) | 52.73 | #101 (GPT-5.1 (2025-11-13)) |
| 14 | Gemini 3 Pro (Preview) (High) | 52.2 | #66 (Gemini 3 Pro (Preview)) |
| 15 | Claude Haiku 5.5 (Max) | 51.54 | #40 (Claude Haiku 5.5) |
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-medcode · How It Works · Data refreshed daily, snapshot 2026-10-20.